Industrial Image Processing

Abstract

Dimensional or shape checking and gauging is one of the most demanding applications of industrial image processing, algorithmically as well as with regard to systems engineering and facility construction. It is actually possible to reach accuracies of just a few light wavelengths, but this requires considerable effort. As in every technical discipline, precise results cannot be achieved without corre-sponding diligence, especially with regard to peripherals, selection of components, mechanical setup, illumination, and image capture. Quality lost in the sensory chain is lost forever. For this reason, an overview chapter on illumination and image capturing will immediately follow this chapter. Dimensional checking can take many forms, from relatively coarse measuring of part dimensions to high accuracy gauging in the micrometer range. Measuring can be an inspection task of its own, e.g. checking whether a part stays within metrically defined allowances, but it can also be a tool for assembly checking. In this case, the precise value of a measurement may be less important than rela-tionships between measurements which give information about the correct position and assembly of a component in relation to other components. Methods and systems developed for optical measuring are correspondingly varied. They range from the determination of relative measurements on the level of image pixels and subpixel-precise gauging as part of standard systems to highly specialized coordinate measuring systems, complete with mechanics, optics and illumination. Such systems can reach accuracies of 1 lm—even more accurate for integral measurements such as an object's centroid—but due to their design they are only applicable for laboratory use or in automated manufacturing lines in a controlled environment (e.g. clean rooms). Before we start to give a feel for the possibilities and prerequisites of this technology, we need to introduce some general terms and concepts. C. Demant et al., Industrial Image Processing, DOI: 10.1007/978-3-642-33905-9_7, Ó Springer-Verlag Berlin Heidelberg 2013 173 7.1 Gauging Tasks Gauging can serve various different purposes and, correspondingly, makes diverse demands on algorithms and sensors. Typical applications with very different requirements are: Assembly checking: Here, gauging methods are used to check the presence of components or other features (like bore holes). Gauging thus becomes a tool for presence verification. Demands on accuracy are typically not very high in this area so that simple pixel measuring methods can often be applied. Shape checking: The purpose of this type of application is to detect deviations from a predefined ideal shape. Some examples are checking whether the edge of a part is sufficiently straight and smooth, whether all pins of an IC are situated on a straight line, whether a group of bore holes is not only present but positioned in an exact circle. This kind of application often requires a high relative accuracy but not a high absolute accuracy. This means that measuring results have to be precisely repeatable and lie within narrow tolerances. A highly accurate calibration for the conversion of pixels to metrical values is less important. Dimensional checking: The most demanding tasks are in the area of actual dimensional checking where certain measurements have to be determined with high absolute accuracy. This not only puts high demands on the measurement computation itself, which will practically always require subpixel interpolation methods, but also on the sensory chain, which has to guarantee absolute stable and repeatable imaging conditions. Only then a calibration is possible that allows for an accurate conversion of image measurements into real values. In the following sections we will provide examples of these types of gauging problems. With each example we will introduce the typical methods and algo-rithms one after the other: pixel measuring, gauging with respect to ideal shapes, and subpixel accurate gauging. Calibration and sensors are discussed in a separate section at the end of the chapter.